Accurate wind forecasts are required to best integrate wind electric potential into scheduling and dispatch decisions.

Wind is difficult to forecast as it is affected by factors such as topography, ground cover and temperature inversions.

The task gets more challenging considering that wind turbines are between 200 to 400 feet above the ground and arrayed in tightly clustered wind farms. Winds at these heights are usually far stronger than at 33 feet, the height used by ground-level weather stations.

A couple of recent developments in the US signify the emphasis being laid on accurate, high-resolution weather forecasts.

The Argonne National Laboratory (ANL) of the network of Laboratories of the US Department of Energy has requested INESC Porto to develop a platform to forecast wind power in the USA. The platform for wind power prediction that INESC Porto is proposing will be linked to a decision support methodology for network operators that will help reduce wind power production and system operation costs. The project started last December and will end in September 2010.

Errors in wind power predictions “can have more severe consequences in the USA than in any European country” due to the country’s temperate and subtropical climate and local geography, said Vladimiro Miranda, director of INESC Porto and coordinator of the project.

In another development, The National Center for Atmospheric Research recently reached an agreement with Xcel Energy to provide detailed, localised weather forecasts that will enable the utility to better integrate electricity generated from wind into the power grid. The forecasts will help operators make critical decisions about powering down coal-fired and natural gas-fired plants when sufficient winds are predicted to generate the required amount of electricity to serve Xcel customers.

William Mahoney, the NCAR programme director overseeing the project, said one of the major obstacles that has prevented more widespread use of wind energy “is the difficulty in predicting when and how strongly the wind will blow at the wind farms”.

But Michael Brower, chief technical officer, AWS Truewind, LLC doesn’t  believe that the difficulty of predicting the wind has “prevented” the widespread use of wind energy.

“Wind is already at one percent of US generating capacity, and in some states and regions it exceeds 10 percent - and its share is growing fast. However, it is certainly one of the factors that have given power companies pause,” said Brower, who is scheduled to speak during Wind Energy Operations and Maintenance Summit USA, to be held on April 1-2 in Dallas, Texas this year.

According to Brower, studies have shown that at low levels of wind penetration (below, say, two-three percent of the local system capacity), variations in output by wind power plants have little noticeable impact on grid operations.

They are in effect “lost in the noise” of demand fluctuations. As the penetration grows, wind energy places increasing demands on load following and other ancillary services required to maintain reliable power delivery to customers. It can also create challenges for managing transmission and distribution networks. The direct costs of these added services (in the absence of wind foresting) have been shown to amount typically to a few dollars per megawatt-hour, or maybe 5-10 percent of the value of the wind generation.

“Wind forecasting is probably the key strategy for reducing the costs associated with wind fluctuations. State-of-the-art forecasting systems can reduce errors in predicted wind generation by half or more compared to no forecast, and this leads directly to lower system operating costs. For the next-day time horizon, the main benefit lies in improving the scheduling of slow-start baseload generation (typically coal and nuclear plants). For the next-hour horizon, forecasting reduces the number of fast-response units that must be on hand to respond to fluctuating net load,” said Brower.

Progress in predictive forecasting techniques

In case of NCAR and Xcel Energy, under the agreement, NCAR will develop a prototype advanced wind prediction system during the next 18 months and will begin to generate test forecasts for Xcel Energy wind farms in Colorado, Minnesota, New Mexico, Texas, and Wyoming after six months. NCAR will continue to improve the system over the following 12 months. Then the prototpe forecasting system will be transferred to Xcel Energy for operational use, while NCAR continues to work toward making the forecasts still more accurate.

According to NCAR, researchers have built a unique modeling system, called Real-Time Four Dimensional Data Assimilation (RTFDDA), that is based on the Weather Research and Forecasting model (WRF). RTFDDA collects diverse weather observations from various platforms (upper-air and surface reports, commercial aircraft reports, mesonet data, and readings from radars, wind profilers, satellites, and other instruments) to provide regional weather analyses, nowcasts, and short-term forecasts.

According to AWS Truewind’s Brower, wind forecasting is merely a branch of weather forecasting, which of course has been around for many decades. “Thus it wasn’t necessary (fortunately) to start from scratch. However, the requirements of wind forecasting are considerably more demanding than those of ordinary weather forecasting. If Accuweather predicts 2 inches of rainfall tomorrow and we get 3 inches instead, nobody is too concerned. But a 50 percent error like that, when applied to wind speed, can create big problems for a utility system with large amounts of wind generation. This demand for accuracy is what has been pushing the technology forward,” said Brower.  

“We have made a great deal of progress. Most of this progress has been the area of better observations, better statistical models (which correct for errors in weather forecasting models), regime-based forecasting (where forecasts are tuned to specific weather conditions), and ensemble forecasting (the use of many different models and data sets to reduce the impact of errors caused by any single model or data set),”  he added.

Unfortunately, Brower says, most of these strategies raise the cost of producing forecasts – but compared to the value forecasting brings, it is a price well worth paying.

“Our estimate is that state-of-the-art forecasts offer a benefit-cost ratio of more than 100 to 1 for most utility systems,” he says.

Wind Energy Operations and Maintenance Summit USA

A session, titled `Why predictive forecasting can dramatically improve your maintenance strategies’, will be conducted as part of Wind Energy Operations and Maintenance Summit USA, to be held on April 1-2 in Dallas, Texas this year.

For more information, click here: http://www.windenergyupdate.com/omusa/programme.shtml

Or

Contact: Tom Evans by email tom@windenergyupdate.com